Unexpected Low DNA Methylation in Transposable Elements at the 5′-CCGG Sites in Three Fruit Tree Cultivars

Abstract

DNA methylation of three cultivars, each of the fruit tree species pear, plum and apple, was analyzed by the methylation-sensitive amplified polymorphism (MSAP) marker. All three fruit tree cultivars were found to contain apparently lower levels of methylation at the 5′-CCGG sites than all other plant species, such as rice and wheat, studied by the same method. Sequencing of the representative loci isolated from the MSAP profiles indicated that both protein-coding genes and transposable elements (TEs) were involved in low methylation. Gel blotting using isolated MSAP fragments and fragment mixtures representing two major types of TEs (copia- and gypsy-like) as hybridization probes confirmed the unexpected low DNA methylation levels at the 5′-CCGG sites in these three fruit tree genomes. Our results suggest that the three asexually propagated perennial fruit trees may indeed contain unusual lower levels of DNA methylation, especially in TEs at the 5′-CCGG sites. Additionally, our results may also suggest that the often used MSAP marker, which targets only one kind of specific methylation-sensitive sites recognized by a pair of isoschizomers (e.g., 5′-CCGG by HpaII/MspI), is not always representative of other cytosine sites (e.g., CHH) or CG sites other than those of 5′-CCGGs in some plant species.

1. Introduction

Covalent modification of cytosine is a major epigenetic marker of eukaryotic nucleus DNA, which plays essential roles via the regulation of genome expression in diverse fundamental biological processes, such as growth and development, genome stability, X-chromosome inactivation, imprinting and response to environmental cues [1,2,3]. Consequently, DNA methylation is essential for normal growth, development and reproduction of all organisms that harbor this epigenetic modification. For example, the loss-of-function mutation in genes coding for DNA methylases that are essential for the maintenance of cytosine methylation results in embryonic lethality in mammals [4]. Likewise, mutants of the DNA methyltransferase 1 gene (met1) in the model plant Arabidopsis thaliana cause severe pleiotropic developmental abnormalities, including sharply reduced fertility and fecundity [5], while the null mutant of this gene (OsMet1-2) in rice causes necrotic death at the seedling stage [6]. In addition, the loss of cytosine methylation in the CG context in rice tissue culture leads to massive mobilization of several otherwise dormant transposable elements (TEs) and can jeopardize genome stability [7].

Although DNA methylation predominately occurs in the symmetrical CG sequence context in mammals, it can happen in virtually all cytosines in plants, including both the symmetric CG, CHG, and asymmetric CHH contexts [2]. DNA methylation has been documented as particularly abundant in plant genomes, due to their often high content of TEs and their derivatives that are heavily methylated in cytosines in all contexts. Thus, a generic feature of plant genomes is that TEs are much more methylated than protein-coding genes [8].

Here, we report that when we were analyzing DNA methylation of three cultivars each of fruit tree species, pear (Pyrus ussuriensis Maxim.), plum (Prunus Salicina Lindl.) and apple (Malus domestica Borkh.), by using the methylation-sensitive amplified polymorphism (MSAP) marker, we unexpectedly found that all three tree cultivars contain apparently lower levels of methylation at the 5′-CCGG sites than other plant species studied by the same method [9]. Moreover, we show that in all the three trees, many protein-coding genes and TEs possess similar levels of methylation at the 5′-CCGG sites, that is, these three fruit tree cultivars do not show markedly differential methylation between genes and TEs as in other plants at this specific type of CG and CHG sites (5′-CCGG) recognized by a pair of isoschizomers, HpaII and MspI. We discuss the implications of our unexpected findings and potential limitations of the MSAP marker that is unheeded in all prior studies.

2. Materials and Methods

2.1. Plant Material

Scions of three fruit tree species, including wild pear (Pyrus ussuriensis Maxim, scion cv. Cold-Delicious/root-stock Pyrus ussuriensis Maxim.), plum (Prunus Salicina Lindl., scion cv. Northern-Red/root-stock Prunus sibirica L.) and apple (Malus domestica Borkh., scion cv. Golden-Red/root-stock Calophyllum inophyllum L.) were used in this study. All three fruit tree cultivars were ca. 10-years-old since grafting, which were produced by the Institute of Fruit Tree Research, Jilin Academy of Agricultural Sciences, and grown at its Experimental Orchards, located at Gongzhuling (N43.5/E124.8), Jilin Province, Northestern China.

2.2. MSAP Analysis

Genomic DNA was isolated from fully expanded young leaves at the same growth and developmental stage of the three scion cultivars growing under the same orchard by the CTAB method [10], and further purified by phenol extractions. The methylation-sensitive amplified polymorphism (MSAP) marker [11] was used to analyze methylation status at globally sampled 5′-CCGG sites. In total, 1 pair of pre-selective primers and 24–48 pairs of selective primers were used for each of the three tree cultivars Table S1. Silver-stained acrylamide gel was used to separate and visualize the amplification products. Only well-discernibale and reproducible bands that present in two independent PCR amplifications (starting from the digestion-ligation step, i.e., the first step of MSAP) were scored [11].

2.3. Sequencing of Selected MSAP Bands

Representative bands (denoting variable DNA methylation at the inner or outer cytosines of the 5′-CCGG sites) in the silver-stained MSAP gels were extracted and re-amplified with the corresponding selective primer combinations. Sizes of the PCR products were confirmed by agarose gel electrophoresis, and cloned into the AT cloning vector (the Sangong Biotech. Inc. Shanghai, China). The DNA segments were Sanger-sequenced with vector primers. The Advanced BlastN and BlastX programs from the NCBI website (http://www.ncbi.nlm.nih.gov/ (accessed on 20 July 2021)) were used for mapping and homology analysis of the cloned DNA sequences gave quality-reads, respectively. For sequenced fragments of the apple tree, they were also mapped to the sequenced genome [12].

2.4. Southern Blot Analysis

Genomic DNA was digested by the methylation-sensitive isoschizomers, HpaII and MspI (New England Biolabs (Beijing, China) LTD). An excess amount of enzymes (10 units of enzyme per µg DNA) and prolonged incubation time (48 h) were employed to ensure complete digestion [7]. Digested DNA was fractionated by 1% agarose gels and transferred onto Hybond N+ nylon membranes, as recommended by the supplier (Amasia Biotechnology Co., Ltd., (Shanghai, China)). Cloned DNA segments, representing methylated sites in either transposable elements (TEs) or protein-coding genes according to homology analysis, were selected as hybridization probes. In addition, the copia-like and gypsy-like mixtures were isolated by polymerase chain amplification (PCR) at an annealing temperature of 37 °C with two pairs of degenerate primers, which are conserved in the reverse transcriptase coding regions of all the copia-like and all the gypsy-like LTR (long-terminal repeat) retrotransposons, respectively. The hybridization signal was detected by the Gene Images CDP-Star detection module (Amersham Pharmacia Biotech, Piscataway, NJ, USA) after washing at a stringency of 0.2 × SSC, 0.1% SDS for 2 × 50 min. The filters were exposed to X-ray films for signal detection.

3. Results

3.1. Overall Levels of Methylated Cytosines at the 5′-CCGG Sites in the Three Fruit Tree Genomes Are Lower Than Expected Based on the MSAP Profiles

The DNA methylation-sensitive amplified polymorphism (MSAP) marker is a modified version of the standard AFLP marker [13] by using methylation-sensitive restriction enzymes [11]. The most often used enzymes in MSAP is a pair of isoschizomers, HpaII/MspI, which are differentially sensitive to cytosine methylation at the 5′-CCGG sites that are known to occur frequently in eukaryotic genomes. Specifically, HpaII will not digest if either or both of the cytosines at a given 5′-CCGG site is/are fully-methylated, but it digests if the external cytosine is hemi-methylated. In contrast, MspI will not cut a 5′-CCGG site if its external cytosine is fully- or hemi-methylated, but it will cut if the internal cytosine is methylated. Thus, the differential methylation states of the cytosines of the 5′-CCGG sites will determine the digestibility of the sites by either of the pair of isoschizomers, and will be reflected in MSAP as the presence (digestible) vs. absence (non-digestible) of amplified bands. MSAP has been documented as an efficient and widely used method to assay methylation status at the 5′-CCGG sites genome-wide [9] before the availability of bisulfite-sequencing-based methylomes for a given organism, or in the field of ecological epigenetics, which entails the interrogation of a large number of samples at the population-level [14].

Using MSAP, we analyzed the methylation states at randomly sampled 5′-CCGG sites of leaf tissue of the three scion cultivars each of fruit tree species, pear, plum and apple. To minimize potential artifacts during the two rounds of PCR amplifications, MSAP was performed twice for all three cultivars starting from the very first step of the technique, i.e., restriction and ligation, and only well discernible and reproducible bands between the two replicates were scored. We found that >95 of the amplified bands displayed in the silver-stained acrylamide gels were reproducible between the two technical replicates.

For pear(scion cv. Cold-Delicious), we used 41 primer combinations and detected a total of 2542 5′-CCGG loci (

Table 1. Number of CCGG loci analyzed by MSAP in the three fruit tree scion cultivars.

Fruit Tree Cultivars	Total 5′-CCGG Loci Scored	Non-Methylated 5′-CCGG Sites	Methylated 5′-CCGG Sites
			CG Fully Methylated	CHG Hemi-Methylated	Total
Pear (scion cv. Cold-Delicious)	2542	2404 (94.6%)	76 (3.0%)	62 (2.4%)	138 (5.4%)
Plum (scion cv. Northern-Red)	2700	2496 (92.5%)	139 (5.1%)	65 (2.4%)	204 (7.5%)
Apple (scion cv. Golden-Red)	1688	1535 (91.0%)	83 (4.9%)	70 (4.1%)	153 (9.0%)

). Of these, fully methylated internal cytosines (e.g., Figure 1a) and hemi-methylated external cytosines (e.g., Figure 1b) were 76 (3.0%) and 62 (2.4%), respectively; the remaining 2404 loci (94.6%) were unmethylated. Hence, the collective methylation level at the 5′-CCGG sites of the pear genome was 5.4% (Table 1). For plum (scion cv. Northern-Red), we used 48 primer combinations and detected a total of 2700 5′-CCGG loci (Table 1). Of these, the full-methylated internal and hemi-methylated external cytosines (e.g., Figure 1c) were 139 (5.1%) and 65 (2.4%), respectively; the remaining 2496 loci (92.5%) were unmethylated. Thus, the total methylation level of the plum genome at the 5′-CCGG sites was 7.5% (Table 1). For apple (scion cv. Golden-Red), we used 24 primer combinations and detected a total of 1688 5′-CCGG loci. Of these, the fully methylated internal and hemi-methylated external cytosines (e.g., Figure 1d) were 83 (4.9%) and 70 (4.1%), respectively; the remaining 1535 loci (91%) were unmethylated. Thus, the overall methylation level at the 5′-CCGG sites of the apple genome was 9.0% (Table 1). Together, based on the MSAP data, the total methylation levels at the 5′-CCGG sites in the genomes of all the three fruit tree cultivars (pear, plum and apple) were 5.4%, 7.5% and 9.0%, respectively, which were strikingly lower than other plant species assayed by the same method [9], including studies from our lab using exactly the same protocol [11,15].

3.2. Classification of a Subset of Genomic Loci Methylated at the 5′-CCGG Sites

To characterize the genomic loci that were methylated at the 5′-CCGG sites from the three studied fruit tree cultivars, we isolated and Sanger sequenced a total of 75 methylated fragments from the MSAP profiles that displayed polymorphism between digests by the pair of isoschizomers, HpaII and MspI (Table S2). The methylated fragments can be divided into four types based on their potential functions inferred from a BlastX analysis. These included known-function genes, putative protein-coding genes, TEs and those with no similarity to the database (

Table 2. Classification of cloned DNA fragments showing alteration in DNA methylation pattern in the fruit tree cultivars based on the MSAP profile.

Category	No. DNA Segment	Percentage (%)
Known-function gene/putative protein-coding gene	26	34.7
TEs	5	6.7
No similarity	44	58.7
Total	75	100

and Table S3) or to annotated genes or TEs in the three sequenced fruit tree genomes, apple [12], plum [16] and pear [17]. Of the three fruit tree cultivars, MSAP fragments with no similarity were 58.7% (Table 2). The known function genes and putative protein-coding genes together were 34.7%, as is detailed below. TEs are often very abundant in plant genomes, accounting for over 50% of the nuclear DNA in many plants with large and complex genomes [8]; however, we found that only 6.7% of the methylated fragments at the 5′-CCGG sites in the fruit trees were TEs (Table 2), indicating that these elements are particularly non- or under-methylated in genomes of the three fruit tree cultivars at this specific type (5′-CCGG) of CG and CHG sites.

3.3. Validation of Methylation States at the 5′-CCGG Sites of Genic Sequences by Gel Blotting

To confirm the methylation states at the 5′-CCGG sites in the three fruit tree cultivars revealed by MSAP, Southern blotting was performed. First, we used 11 isolated MSAP fragments that are known or hypothetic genic sequences as probes. Two of these fragments (MSAP 50 and MSAP 51) were isolated from the pear cultivar. The proteins encoded by these two fragments include homologues to a beta-galactosidase-complementation protein (MSAP50), and a beta-galactosidase alpha-peptide (MSAP 51) (Table S2). Five fragments (MSAP6, MSAP13, MSAP19, MSAP66 and MSAP71) were isolated from the plum cultivar. The proteins encoded by these five fragments are homologues to a midasin-related protein (MSAP6), a fiber protein Fb2 (MSAP13), an expressed protein (MSAP19), an unnamed protein product (MSAP66), and a callose synthase 1 (CALS1)/1,3-beta-glucan synthase 1 (MSAP71). Four fragments (MSAP53, MSAP58, MSAP84 and MSAP88) were isolated from the apple cultivar. The proteins encoded by these four fragments included homologues to a 1-deoxy-D-xylulose-5-phosphate reductoisomerase (MSAP53), a PDR-type ABC transporter 1 (MSAP58), a hypothetical protein (MSAP84), and a TdcA1-ORF2 (MSAP88). According to the MSAP profiles, of these eleven fragments, four showed hemi-methylation of external cytosines (MSAP13, 19, 53, and 84) and seven showed full methylation of internal cytosines (MSAP6, MSAP66, MSAP71, MSAP50, MSAP51, MSAP58, and MSAP88) at the 5′-CCGG sites. As exemplified in Figure 2, the gel blotting patterns by these genic fragments as hybridization probes largely recapitulated the differential methylation states between the inner and outer cytosines at the 5′-CCGG sites based the MSAP pattern categorizations in the respective fruit tree cultivars (Table S2).

3.4. Low Level DNA Methylation of TEs at the 5′-CCGG Sites Detected by DNA Gel-Blotting Hybridization

TEs and their derivatives are ubiquitous and abundant components of plant genomes, which often make up over 50% of the nuclear DNA. TEs are generally silenced by heavy cytosine methylation of all sequence contexts in plants [8]. However, in the three fruit tree cultivars we studied here, the TEs only constituted 6.7% of the methylated fragments (Table 2), indicating that these TEs are hypomethylated at the 5′-CCGG sites in the genomes of the three fruit tree cultivars. To further investigate the level of DNA methylation of TEs at the 5′-CCGG sites by an independent approach, we selected five DNA fragments derived from TEs as probes to perform methylation sensitive gel-blotting hybridization, as was the case for genic sequences, described in the foregoing section. Of these five fragments, one was isolated from the pear cultivar (MSAP39), three were isolated from the plum cultivar (MSAP 2, MSAP3 and MSAP76), and one was isolated from the apple cultivar (MSAP61). As shown in Figure 3, all five selected TEs were highly hypomethylated at the 5′-CCGG sites in these fruit tree genomes, and hence, validated the MSAP results.

Based on modes of transposition, TEs are divided into two major classes, i.e., DNA transposons and retrotransposons [8]. Retrotransposons are divided into two types based on their structures, including those that are flanked by long terminal repeats (LTRs), and those that do not contain LTRs. LTR retrotransposons can be further categorized into two major families, copia-like and gypsy-like. The main structural difference between copia-like and gypsy-like families lies in the order of the reverse transcriptase (RT) and integrase domains in their pol genes [8]. Because the copia-like and gypsy-like retrotransposons are the most abundant TEs in higher plant genomes [8], we further interrogated their collective methylation states at the 5′-CCGG sites by gel blotting hybridization. Specifically, methylation-sensitive gel blotting was performed by using the mixed copia-like and gypsy-like fragments as probes, respectively, which were amplified by degenerate primers that targeted the conserved reverse transcriptase (RT) region of each type of elements, respectively. By this strategy, a large amount of TE fragment copies of each family could be detected. The gel blotting results showed collective low levels of DNA methylation of both families of element mixtures in all three fruit tree genomes at the 5′-CCGG sites (Figure 4), similar to the levels detected by the five individual TEs (Figure 3) and the genic sequences (Figure 2). Taken together, it is clear that many TEs of all the three fruit tree cultivars we studied here (wild pear, plum and apple) are hypomethylated at the 5′-CCGG sites.

4. Discussion

Despite the near ubiquitous occurrence of DNA methylation in virtually all higher eukaryotes, the extent of this covalent modification is known to vary markedly across biological taxa. Nevertheless, higher plants (flowering plants) usually possess higher DNA methylation levels than most mammalian animals, due to the often larger and more complex genomes of plants [1]. Thus, our finding that the three studied horticulturally important fruit tree cultivars of pear (Pyrus ussuriensis Maxim.), plum (Prunus Salicina Lindl.) and apple (Malus domestica Borkh.) all have very low levels of DNA methylation at the large number of 5′-CCGG sites sampled genome-wide by the MSAP marker is unexpected. Even more surprising, we found that TEs of these three fruit tree cultivars are not more methylated than the protein coding genes at these specific sites. We can rule out the possibility that our results are due to artifacts of the PCR-based MSAP marker because the patterns of DNA-gel blotting using the 11 isolated MSAP fragments that are genic and five fragments that are TEs as probes have independently validated the MSAP results. Moreover, when the copia-like and gypsy-like mixtures (which were amplified by using degenerate oligonucleotide primers, and hence should have sampled the RT regions of many TE members of each of the two major TE families) were used as hybridization probes, the gel-blotting patterns also clearly revealed lower methylation than expected for such elements in plants.

The whole genome bisulfite-sequencing-based methylome of the apple tree was recently reported [12]. It was found that protein-coding genes of the apple genome can be categorized into the following three major clusters: cluster 1 are those that show low DNA methylation of all sequence contexts (CG, CHG and CHH) in both gene bodies per se and their immediate flanks, cluster 2 are those that show low DNA methylation in gene bodies but higher methylation in flanks, and cluster 3 are those that show higher methylation in both gene bodies and flanks [12]. Thus, it is likely that the genic loci we sampled by the MSAP marker in the apple cultivar have been preferentially targeted to the first two clusters of genes, conceivably due to their higher openness in local chromatin states [18], and hence more accessibility to the isoschizomers used in our MSAP assay. Although methylomes were also reported in other fruit trees, they were for different purposes, such as the influence of DNA methylation on subgenome dominance in pears [19] and between tissue methylation differences in almond [20], and hence, cannot be directly referenced here. With respect to the percent value of TEs, the three fruit tree cultivars we studied were 57.3% in apple [12], 53.1 in pear [17], and 48.3 in plum [16]. Apparently, these three fruit tree cultivars contain genomic proportions of TEs that are similar to many other plant species. In addition, compared with genes, TEs are also more methylated in the apple genome than genes [12], as in other plants. Notably, the methylome analyses indicated that the overall methylation levels of TEs in the apple genome were indeed lower than TEs of most other plant methylomes, such as wheat [21], cotton [22] and rice [6], but they are still higher than those of genes [12]. Thus, to an extent, our MSAP-based methylation data are inconsistent with the methylome. We suspect this discrepancy might be due to the very nature of the MSAP marker, which is exclusively based on differential digestibility by the isoschizomers, HpaII/MspI [9]. Therefore, we were only able to assess methylation states of the 5′-CCGG sites but no other cytosines. As a result, the unusual low methylation we detected for both genes and TEs, but especially for TEs, might be specifically confined to the 5′-CCGG sites rather than cytosines of all CGs or cytosines of other contexts. Another possibility is that the scion cultivars we used have been particularly affected in DNA methylation by the different rootstocks, given that grafting can indeed cause changes in DNA methylation [23]. This possibility, however, is unlikely as the methylation changes induced by grafting were found to be locus-specific and include both hypo- and hypermethylation, thus rendering little change in the overall methylation level [23]. It is also unlikely that some unknown environmental factors induced hypomethyaltion in these cultivars, as all three fruit tree cultivars were grown for ca. 10 years in a usual orchid (Materials and Methods) for the commercial purpose of propagating cutting branches. In any case, apart from intrinsic biological reasons (most likely), our results draw the conclusion that methylation states of the cytosine bases of the 5′-CCGG sites probably cannot represent those of the other CG sites and cytosines of other contexts, at least in the apple genome. Due to either lack of methylomes or directly comparable methylome information of the other two fruit trees (pear and plum) we studied here by MSAP, it is currently unknown whether a similar situation will appear in these two fruit tree cultivars. Notably, however, our MSAP-based methylation level estimates at the 5′-CCGG sites indicated that apple tree was the highest (9.0%) while those of pear and plum were even lower, being 5.4% and 7.5, respectively. Therefore, it is reasonable to speculate that a similar inconsistency between the methylation level at the 5′-CCGG sites and at other sites or contexts likely holds for these two fruit tree genomes as well.

Recent studies have shown that DNA methylation undergoes dynamic changes in the process of fruit development and ripening, including those of fruit trees. For example, fruit development and ripening in both sweet orange (Citrus scinensis L.) [24] and rape (Vitis vinifera L.) [25] was accompanied by increased DNA methylation. We note that our sampling was restricted to one tissue at a single developmental stage (fully expanded young leaf); thus, we cannot fully rule out whether the results may vary across development and in different tissues/organs. Although based on prior findings in other plants, this possibility is unlikely within vegetative tissues [3,6] and it is possible when reproductive tissues are included. For example, a recent study in almond (Prunus dulcis (Mill.) D.A. Webb.) found clear variation in genome-wide DNA methylation values across six tissues (leaf, flower, exocarp, mesocarp, endocarp, and seed coat), with the leaf exhibiting the lowest value and flower the highest [20]. Thus, it will be interesting to analyze other tissues of these fruit tree cultivars, and to investigate whether the phenomena we found in the leaf tissue will hold in other tissues, especially those related to reproduction.

At any rate, our observation that TEs are less methylated than low-copy genic sequences in these three fruit trees at the 5′-CCGG sites appears incongruent with the general trend in most other plant and animal species [1,8]. However, we are not the first to find that TEs are less DNA methylated than are genes. A previous study in an invertebrate chordate species (Ciona intestinalis) showed even more striking atypical patterns of DNA methylation between TEs and genes; it was documented in that study that TEs are nonmethylated, while genes are predominantly methylated in C. intestinalis [26]. Together, it is clear that exceptions to general patterns and paradigms do happen with respect to DNA methylation, as it happens in many other aspects of biology.

5. Conclusions

The methylation states at the 5′-CCGG sites in leaf tissue of the scion cultivar genomes of three fruit trees (pear, plum and apple) detected by the MSAP marker were extremely low. This result was confirmed by DNA Southern blot analysis using 11 isolated MSAP fragments as probes. Furthermore, methylated transposable elements are much lower in proportion than expected from the genome composition. Together, our results suggest that epigenetic modification in the form of DNA methylation can be dramatically diverse in plant species, and is probably related to their life history traits, such as sexual vs. asexual reproduction models and annual vs. perennial life cycles, or still unknown factors.